Chapter II: Hidden Donors: Analyzing the Censoring Problem in U.S. Federal
2.6 Differences in Contribution Patterns
Figure 2.2: Distribution of Contribution Sum to Sanders, Visible and Hidden Con- tributors, Truncated at $500
0 20000 40000
0 50 100 150 200 250 300 350 400 450 500
Contribution Sum to Sanders (USD)
Frequency
Donor Visibility Hidden Visible
average donation size to the campaign was $27, and thus the campaign relied on small donors and not on the fat-cats of Wall Street.18 This emphasis has been frequently covered by the media (Bump, 2016; Foran, 2016; Mehta et al., 2016), and was further echoed in Sanders’s later campaign strategies, as they added a $27 donation option in their suggested contribution amounts. And indeed, the campaign had successfully mobilized small donors enough that its median donor was a one-time giver of $27. However, $27-giving is more indicative of a visible donor compared to the benchmark distribution. While 20% of Sanders donors have given $27 at least once, conditional on a $27 giving record, a donor is 26.5% likely to be visible. This may suggest that such compliance with the campaign message—enough to choose such a non-salient amount—is indicative of donor loyalty. That is, who ends up contributing enough to earn the visibility.
Aside from the obvious differences in contribution amounts, Table2.3 also shows important differences by the number of contributions. Note that visible donors have
18Note that the individual contribution limit of 2016 was $2,700, which could have resulted in this value.
given multiple times—indeed, while the median hidden donor has given only once, the median visible donor has given ten times to Sanders. This shows that not all visible donors are one-time donors of hundreds of dollars.
Timing
The timing of campaign contributions has not been widely analyzed, though there are a few recent papers on the subject (Christenson and Smidt, 2011; Smidt and Christenson, 2012; Mitchell et al., 2015; Magleby et al., 2018). These studies aggregate the number of donations on a daily level to study donation dynamics.
Magleby et al.(2018) aggregated the data separately for small and large donors, and concluded that the same events motivate large and small donors alike, and thus they seem to be driven by the same dynamics.
However, the timing of giving disaggregated by first and last date of contributions may provide additional insight into donor motivations. If we view contributing to campaigns as a rational act in which a donor weighs her marginal benefits and costs, whether her gift is expressively or instrumentally motivated, the first contribution date reveals when the marginal benefits first exceed marginal costs of the person. If donations are driven by solicitation, timing may show how an equilibrium is struck between demand and supply.
Table2.3also shows various summaries of when the donors choose to give. Notice that visible contributors on average enter earlier in the race with their first contribu- tion. While visible donors give long before the primaries, hidden donors enter right before the primaries start. Visible donors also leave later than hidden contributors, again by a difference of two to three months, a substantial amount of time. Note how half of the hidden donors started to leave after March 9th, although that was the day after the Sanders’ surprise win in Michigan. In fact, by April 26 (the time of the so-called Acela primaries), less than one in six Sanders hidden donors remained (16.9%). For visible donors, a majority of them stopped only after May 18.
Table2.3shows the percentage of donors who gave to Sanders or other committees aftersome key dates in which Sanders’ prospect declined. We include Super Tuesday (March 15), the Acela primaries (April 26), when Clinton became a presumptive nominee (June 6), when Sanders officially endorsed Clinton (July 12th), and when the Democratic Convention officially closed (July 28). Note that the percentage of hidden donors who give to Sanders rapidly declines as the election progresses.
This contrasts with visible donors. By the time that Clinton is the presumptive
nominee, almost 30% of visible donors still opened their wallets for further campaign contributions to Sanders, while only 5% of hidden donors did.19 This suggests that “hidden” status was achieved because they were not attracted to give until the race was contested, and they were also quicker to lose interest. That hidden donors are late entrants to the market signals that they may not have been strong supporters to begin with, and that they may have been persuaded to give only after Sanders gained sufficient momentum.20 Indeed, it is surprising to see that so many visible donors contributed to Sanders after he had effectively lost his quest for the Democratic nomination. Even if we account for the possibility of campaigns mislabeling donations with different dates then they actually received, the percentage seems substantial.
The difference in enthusiasm could be attributed to many factors, such as ideological differences (Ensley, 2009), candidate valence, information costs, or contributor’s budgets. For example, given that donors are instrumental, the ones who are further away from Sanders ideologically may have been attracted to give by his rising prospects, but as his viability fell, they withdrew as their marginal benefit of giving fell below the costs. It could be that these are donors who are generally not sufficiently interested in politics to consider giving, and only when Sanders became a serious figure, were attracted just enough to give once or twice. There could be a donor class with limited budget for political donations, for instance due to a lower family income, who thus cannot afford to give expressive donations such as those made after Sanders loses the primary race. With only administrative data we cannot immediately conclude whether either of these two explanations hold true, but this is worthy of further investigation.
Differentiating Visible Donors
Given the results above, there is a strong possibility that visible donors become vis- ible simply because they have patiently given many times, and eventually gave more than the $200 threshold. These donors, who we calleventually visibledonors, may be different from the ones who have given over $200 in their first contribution, who
19Overall, 22.2% of donors started to give before the Iowa Caucuses but also stopped before the Iowa Caucuses, while 35.2% of donors gave before and after the Iowa Caucuses. Similarly, 35.9%
started to give before Super Tuesday but not after, while 39.1% of donors gave before and after Super Tuesday.
20Note that this difference in timing is best highlighted also because Sanders had a lengthy campaign, as he did not leave the race until the Democratic convention. If a less-viable candidate had dropped out earlier or his/her prospects dwindled much more quickly, while the difference will still hold, it would be more subtle.
we will callimmediately visible donors. If for instance a donor gives $50 monthly starting January 2016 and opts out during or before April, she will be hidden, while if she remains to give once more until May, she becomes visible. Distinguishing the different types of visible donors may be important in understanding differences in their political behavior from hidden donors.
Table 2.4 replicates Table2.3 with these classifications. The immediately visible donors constitute only 2.7% of the donor population, while the eventually visible donors are 9.7% of the donor population. Note that the number and sum of contribu- tions are correlated, as there is a $2,700 cap on the maximum amount of individual contributions—a donor who gives $500 regularly cannot give more than 5 times.
Altogether, the immediately visible donors (2.7% of donors) give 20% of the indi- vidual contributions amount, the eventually visible donors (9.7%) give 46.2%, and the hidden donors (87.6%) give 33.8%. Again, this now excludes the first $200 from visible donors, and hence the proportion is reduced from what is presented in Table 2.1. Each class is a formidable force in the financial electorate, but it is the eventually visible donors who filled the campaign chest of the presidential candidate. Table2.2’s equivalent with these classifications is available in Appendix B.5.
Table 2.4: Sanders Contributions, Immediately Visible, Eventually Visible, and Hidden Donors, 2016 Sanders Campaign
Immediately Visible (2.7%)
Eventually Visible (9.7%)
Hidden (87.6%) Sum of contributions:
—— Mean $876.0 $564.0 $45.7
—— Q1 $277.0 $269.0 $15.0
—— Q2 $500.0 $389.0 $27.0
—— Q3 $1,000.0 $638.0 $60.0
Number of contributions:
—— Mean 6.2 15.1 2.5
—— Q1 1 7 1
—— Q2 3 11 1
—— Q3 7 19 3
Median first contribution date 01/25/2016 09/23/2015 01/29/2016 Median date when voter became visible 01/25/2016 03/04/2016 — Median last contribution date 04/03/2015 05/25/2016 03/09/2016
% of donors who gave to Sanders, after
—— Super Tuesday 61.5% 89.6% 44.9%
—— Acela primaries (Super Tuesday III) 38.7% 70.1% 16.9%
—— Clinton became presumptive nominee 15.7% 33.3% 4.6%
—— Sanders endorsed Clinton 3.0% 6.2% 0.8%
—— End of Democratic convention 0.1% 0.0% 0.0%
On average, the eventually visible donors start early, but give frequently and are late to leave the donation game. For other campaigns than the Sanders campaign, we would not have been able to observe their first few donations, and hence would not know exactly when they entered. Instead, we would have observed only the date that they crossed the $200 threshold and became visible, the median of which is March 3, 2016, and would have only known that they would have entered before that.
Because the best predictor of giving is their past giving history, it makes sense for campaigns to target early givers and attempt to solicit more money from them, given that they have not maxed out. It appears that donors who become eventually visible are the most loyal—or persuadable—set of donors. Indeed, conditional on having first given before June 1, 2015 (within two months of launching the campaign), 27.6% will become eventually visible.
One interesting point is that the median date of first donation for hidden donors and for immediately visible donors is similar—only four days apart, a week to go before the Iowa caucuses. The first and third quartiles are also similar: respectively September 12, 2015 vs. October 1, 2015, and March 6, 2016 vs. March 9, 2016.
If first donation dates are indication of interest/persuasion, this suggests that the hidden and immediately visible donors may be quite similar in their motivations for contributing, and perhaps that income or wealth is the primary difference between hidden and immediately visible donors. It is the repeat donors—who are also mostly early donors—who stand out, and they might be the donors who most affect the campaign’s policy positions.
Again, we are looking at the data from a single campaign. Depending on the nature of different campaigns, contextual factors, and primary election dynamics, the profile of the hidden donor population might vary across election cycles and candidates. If the hidden donors—who provided one-third of contribution dollars to Sanders in 2016—can be further mobilized with certain issue positions, a campaign may choose to do just that. Hidden donors have, after all, the potential to give more and perhaps become visible. If they are more economically representative, the campaign may find it more beneficial to focus on small donors and retaining policy positions closer to the median voter. However, if the campaign’s donor base cannot be mobilized this way, the campaign may choose to diverge from the median voter to attract more rich donors.
Are the three classes of donors actually any different in policy preferences? We cannot conclude that they are with only administrative data, and much less with
just Sanders’ 2016 campaign. However, we believe it is worth further investigation.
Uncovering hidden donors clearly shows that we need to better understand who the presidential campaign perceives to be major financial electorate, and whether catering improves a campaign’s chances of primary and general election success.